[WIP] Add settings to toggle %gp_rel usage on generated assembly

CHANGELOG.md

@@ -1,10 +1,17 @@
 # splat Release Notes
 
+### 0.25.2
+
+* Two new yaml options: `use_gp_rel_macro_nonmatching` and `use_gp_rel_macro`
+  * Allows to toggle the use of `%gp_rel`s.
+  * Not using `%gp_rel`s may be desirable for projects that use old assemblers that do not support said relocation parameter.
+  * The assembler is free to choose the kind of relocation to use if no explicit relocation parameter is used for a given instruction, it may even expand the instruction into multiple instructions. To avoid this, it is the user's responsability to provide the symbol's information (like the symbol's size) to the assembler so it can pick the correct relocation.
+
 ### 0.25.1
 
 * Added two new segment types: `gcc_except_table` and `eh_frame`.
   * Used by GCC to handle C++ exceptions.
-* New yaml option: `asm_jtbl_label_macro`
+* New yaml option: `asm_ehtable_label_macro`
   * Allows to specify the macro used by ehlabels, the ones generated by `gcc_except_table` references.
 
 ### 0.25.0

README.md

@@ -21,7 +21,7 @@ The brackets corresponds to the optional dependencies to install while installin
 If you use a `requirements.txt` file in your repository, then you can add this library with the following line:
 
 ```txt
-splat64[mips]>=0.25.1,<1.0.0
+splat64[mips]>=0.25.2,<1.0.0
 ```
 
 ### Optional dependencies

docs/Configuration.md

@@ -635,6 +635,26 @@ Allow specifying that the global memory range may be larger than what was automa
 
 Useful for projects where splat is used in multiple individual files, meaning the expected global segment may not be properly detected because each instance of splat can't see the info from other files, like in PSX and PSP projects.
 
+### use_gp_rel_macro_nonmatching
+
+If True then use the `%gp_rel` explicit relocation parameter on instructions that use the $gp register, otherwise strip the `%gp_rel` parameter entirely
+and convert those instructions into macro instructions that may not assemble to the original bytes.
+
+In the latter case, it is the user's responsability to provide the symbol's information to the assembler so it can assemble the instruction with the
+proper relocation, for example by declaring the required symbol on the corresponding `.c` or `.cpp` file.
+
+Turning off this setting may be useful for projects with old assemblers that do not support `%gp_rel`, like PS2 and PSP projects.
+
+This setting is applied exclusively to `c` segments (functions under the nonmatchings folder).
+
+Defaults to `True`
+
+### use_gp_rel_macro
+
+Does the same as `use_gp_rel_macro_nonmatching`, except it is only applied to `asm` and `hasm` segments.
+
+Defaults to `True`
+
 ## N64-specific options
 
 ### header_encoding

pyproject.toml

@@ -1,7 +1,7 @@
 [project]
 name = "splat64"
 # Should be synced with src/splat/__init__.py
-version = "0.25.1"
+version = "0.25.2"
 description = "A binary splitting tool to assist with decompilation and modding projects"
 readme = "README.md"
 license = {file = "LICENSE"}
@@ -20,7 +20,7 @@ dependencies = [
 
 [project.optional-dependencies]
 mips = [
-    "spimdisasm>=1.27.0,<2.0.0", # This value should be keep in sync with the version listed on disassembler/spimdisasm_disassembler.py
+    "spimdisasm>=1.28.0,<2.0.0", # This value should be keep in sync with the version listed on disassembler/spimdisasm_disassembler.py
     "rabbitizer>=1.10.0,<2.0.0",
     "pygfxd",
     "n64img>=0.1.4",

requirements.txt

@@ -4,7 +4,7 @@ tqdm
 intervaltree
 colorama
 # This value should be keep in sync with the version listed on disassembler/spimdisasm_disassembler.py and pyproject.toml
-spimdisasm>=1.26.0
+spimdisasm>=1.28.0
 rabbitizer>=1.10.0
 pygfxd
 n64img>=0.1.4

src/splat/__init__.py

@@ -3,7 +3,7 @@
 __package_name__ = __name__
 
 # Should be synced with pyproject.toml
-__version__ = "0.25.0"
+__version__ = "0.25.2"
 __author__ = "ethteck"
 
 from . import util as util

src/splat/disassembler/spimdisasm_disassembler.py

@@ -7,7 +7,7 @@
 
 class SpimdisasmDisassembler(disassembler.Disassembler):
     # This value should be kept in sync with the version listed on requirements.txt and pyproject.toml
-    SPIMDISASM_MIN = (1, 27, 0)
+    SPIMDISASM_MIN = (1, 28, 0)
 
     def configure(self):
         # Configure spimdisasm

src/splat/segtypes/common/c.py

@@ -36,6 +36,8 @@ def __init__(self, *args, **kwargs):
 
         self.file_extension = "c"
 
+        self.use_gp_rel_macro = options.opts.use_gp_rel_macro_nonmatching
+
     @staticmethod
     def strip_c_comments(text):
         def replacer(match):

src/splat/segtypes/common/codesubsegment.py

@@ -41,6 +41,7 @@ def __init__(self, *args, **kwargs):
         )
 
         self.is_hasm = False
+        self.use_gp_rel_macro = options.opts.use_gp_rel_macro
 
     @property
     def needs_symbols(self) -> bool:
@@ -59,6 +60,7 @@ def configure_disassembler_section(
         section.isHandwritten = self.is_hasm
         section.instrCat = self.instr_category
         section.detectRedundantFunctionEnd = self.detect_redundant_function_end
+        section.gpRelHack = not self.use_gp_rel_macro
 
     def scan_code(self, rom_bytes, is_hasm=False):
         self.is_hasm = is_hasm

src/splat/util/options.py

@@ -183,7 +183,7 @@ class SplatOpts:
     asm_data_macro: str
     # Determines the macro used at the end of a function, such as endlabel or .end
     asm_end_label: str
-    # Determines the macro used to declare data symbols in asm files
+    # Determines the macro used to declare ehtable labels in asm files
     asm_ehtable_label_macro: str
     # Toggles the .size directive emitted by the disassembler
     asm_emit_size_directive: Optional[bool]
@@ -226,6 +226,14 @@ class SplatOpts:
     # Useful for projects where splat is used in multiple individual files, meaning the expected global segment may not be properly detected because each instance of splat can't see the info from other files.
     global_vram_start: Optional[int]
     global_vram_end: Optional[int]
+    # For `c` segments (functions under the nonmatchings folder).
+    # If True then use the `%gp_rel` explicit relocation parameter on instructions that use the $gp register,
+    # otherwise strip the `%gp_rel` parameter entirely and convert those instructions into macro instructions that may not assemble to the original
+    # bytes. In the latter case, it is the user's responsability to provide the symbol's information to the assembler so it can assemble the
+    # instruction with the proper relocation.
+    use_gp_rel_macro_nonmatching: bool
+    # Does the same as `use_gp_rel_macro_nonmatching`, except it is only applied to `asm` and `hasm` segments.
+    use_gp_rel_macro: bool
 
     ################################################################################
     # N64-specific options
@@ -550,6 +558,10 @@ def parse_endianness() -> Literal["big", "little"]:
         ),
         global_vram_start=p.parse_optional_opt("global_vram_start", int),
         global_vram_end=p.parse_optional_opt("global_vram_end", int),
+        use_gp_rel_macro_nonmatching=p.parse_opt(
+            "use_gp_rel_macro_nonmatching", bool, True
+        ),
+        use_gp_rel_macro=p.parse_opt("use_gp_rel_macro", bool, True),
     )
     p.check_no_unread_opts()
     return ret